Abstract: In the positioning problem of distributed optical fiber sensor system, traditional time delay estimation (TDE) algorithms usually fail due to strongly correlated noise.A modified general cross correlation method was used to remove the correlated noise. According to the system features, a new TDE scheme was proposed to further improve the distributed optical fiber sensor system performance on accuracy and stability. The received data were first divided into many frames on the basis of event signal and then Kalman filter was applied to track the TDE results of the data frames. Simulation experiments and real data tests indicated that the proposed TDE scheme had good performance to suppress the correlated noise and improve the TDE accuracy and stability. Lots of on-site tests proved the TDE error was stably controlled below the sampling interval of 0.2, which satisfied the system accuracy requirement.